Credit Fraud Detections - ML Approach

This repository contains the train and deploy folder.

/train contains all the notebooks, files and data required to analyze the data, train and tune the model.

/deploy contains the Flask API assets to serve the model as an endpoint and get predictions from it.

To run the Flask API locally you may start having this steps inside a virtual environment:

in deploy/

$ pip install -r requirements

and then run the flask app with

$ flask run

You can start making inferences sending a POST requests as:

import requests

json_data = {"V1": -0.365234375, "V2": 0.1234415820, "V3": 0.52880859375,"V4": 0.055511474609375,"V5": -0.045166015625,
 "V6": -0.043853759765625,"V7": -0.12164306640625,"V8": 0.202880859375,"V9": 0.05010986328125,"V10": -0.279296875,
 "V11": -0.260009765625,"V12": -0.0565185546875,"V13": -0.279296875,"V14": 0.051910400390625,"V15": 0.06829833984375,
 "V16": 0.0885009765625,"V17": -0.042877197265625,"V18": -0.02642822265625,"V19": -0.34423828125,"V20": -0.05657958984375,"V21": 0.04327392578125,
 "V22": 0.056427001953125,"V23": -0.038238525390625,"V24": 0.01045989990234375,"V25": 0.00083160400390625,"V26": -0.132568359375,
 "V27": 0.005035400390625,"V28": -0.00994110107421875,"Amount": 88.67}

api_endpoint ='http://127.0.0.1:5000/predict'
pass_r = requests.post(api_endpoint, json=json_data)

You may find different requirements.txt in each folder. This is because the training and deploy are made to be different environments each.